Long‐term monitoring on a new channelized stream section: changes in mesohabitat, composition, and size structure of fish assemblages

The construction of riverine structures (weirs, bridges, or channelization) on riverbeds causes alterations in the flow regime and channel geometry. Once a new stretch is created, species must colonize it. The ecological succession processes that are entailed are decisive for adequate recovery after alterations, and understanding these processes would enable elaborate efficient restoration actions. We analyzed environmental variables and the colonization and succession patterns of fish assemblages in a new channelized stretch of a river from its construction to the present (from 1996 to 2020) in the northern Iberian Peninsula. During the studied period, habitat diversity increased, and mesohabitat became more complex. Depths, depth diversity, and the number of pools in the new channel increased with time. Water temperature decreased because of the new shade provided by riparian forests. The size‐related variables of the fish community (size diversity, mean, and maximum length) increased in the new section, achieving similar values to those in the control section. The slopes of the fish size spectra showed a slow evolution over 25 years from a fish assemblage dominated by small fish to a more size‐diversity fish community. Our results suggest that habitat complexity shape fish assemblage and structure. Moreover, size‐related variables can be effective ecological indicators of fish colonization and succession processes. Finally, small‐scale restoration measures in the riverbed and riparian forest are expected to increase the effectiveness of future restoration projects in rivers.


Introduction
Freshwater environments are subject to multiple stressors stemming from human activity (Woodward et al. 2010).Rivers suffer direct and indirect human impacts over watercourses and adjacent territories (Vörösmarty et al. 2010).One of these severe alterations of waterways is the construction of riverine structures (dams, bridges, channelization, etc.), causing diversions and changes in the flow course.Newly constructed river sections must then be colonized by the species present in the ecosystems.After the early colonization of the new habitat, a succession process occurs over time, with changes in species' presence and relative abundance.Colonization and succession are essential processes that regulate the persistence of metapopulations and the ecological resilience of river communities (Albanese et al. 2009).The theory of ecological succession has been mostly developed for plants in terrestrial ecosystems (Foster & Tilman 2000;Anderson 2007); however, it has also been evaluated for fish colonization of artificial reefs in the sea (Bohnsack 1991;Brickhill et al. 2005), lakes, and reservoirs (Bassett 1994;Santos et al. 2011).Studies on the colonization and succession processes of fish communities in rivers altered by man-made structures have apparently been neglected.Previous literature has focused on evaluating these processes after droughts or mass mortality events due to pollution (Albanese et al. 2009).Because the succession process is strongly influenced by distinctive habitat properties, restoration projects after the construction of new structures on riverbeds can facilitate population and fish community recovery.To be effective, these restoration projects require knowledge of the environmental and biological elements determining this recovery.
Ecologists have used taxonomic richness and diversity to analyze ecological communities, commonly assuming a positive relationship between these attributes and the ecological status in a complex-mature ecosystem (Woodward et al. 2010).An alternative to the taxonomic approach for assessing these communities is to study the body size structure of the organisms within them (White et al. 2007).Body size is a fundamental attribute of organisms because many physiological rates are size-dependent (Arranz et al. 2021).Body size also determines ecological structures and processes, including population abundance, predator-prey interactions and the resilience capacity of food webs (Arranz et al. 2021).Although studies about fish community size structure have mainly focused on simple sizerelated variables (i.e.mean or maximum length), other more complex approaches (i.e.size spectrum, size diversity) can be used to obtain a more holistic view of the size structure in a fish community.For instance, the abundance size spectrum describes the relationship between body size and abundance.The y-intercept of the linear size spectrum can be interpreted as an approximation of food web capacity (Murry & Farrell 2014).In contrast, the slope can be interpreted as a measure of energy transfer or ecological efficiency (Emmrich et al. 2011).On the other hand, the size diversity index (Quintana et al. 2008) integrates the amplitude of the size range and the evenness across sizes, in the same way as species diversity integrates the number of species and their relative abundance.Thus, high size diversity values appear in cases of equally distributed fish lengths and when the catch consists of many different fish lengths (i.e.large size range; Emmrich et al. 2011).
Various studies on river fish assemblages have shown that these more complex size-related variables have responded to altitudinal gradients in pristine streams (Benejam et al. 2018), as well as to different kinds of anthropogenic pressures, such as changes in land use associated with urbanization, nutrient concentration and the presence of nonnative species (Murry & Farrell 2014;Arranz et al. 2021).In this study, we tested whether size spectra and size diversity can help to evaluate the ecological recovery of river fish communities after an impact.We analyzed the colonization patterns of fish assemblages in a new channelized section starting 1 year after its construction in 1995-2020.Our objective was to address how environmental variables and time mediate fish colonization and succession to understand the mechanisms of these ecological processes in a new aquatic ecosystem.In other words, our study aims to describe fish colonization and succession processes on an artificial structure in a river by assessing fish size assemblage and its variation over a long-term period.Finally, we aim to assess the suitability of size-related variables as indicators of fish population changes in river colonization and succession processes.

Study Area
This study was conducted on the Larraun River (Navarra, northern Spain) in the upper Ebro Basin.The Larraun River runs 21 km from its source in the Aralar Mountains (650 m a.s.l.) to its confluence with the Arakil River in the village of Irutzun (484 m a.s.l.), draining an area of 221 km2 .The hydrological regime has a pluvial oceanic Atlantic climate with a mean annual temperature above 10 C and annual rainfall above 1,000 mm.
As a result of the construction of a mountain highway between 1991 and 1995, the Larraun River suffered different alterations in its middle and lower zones.One of the significant alterations was the channelization and straightening of a 500-mlong section through which the river was diverted, creating a new river stretch.After the construction of the highway (1995), waters of the Larraun River began flowing through a new artificial channelized stream section close to the village of Urritza (42 57 0 60 00 N, 1 49 0 56 00 W).Unfortunately, at that time, legal obligations for restoration did not exist.Consequently, no actions to naturalize the new channelized stream section were performed.
For this study, two sites in the middle Larraun River were analyzed: the new channelized stream section described above, named "Corta de Urritza," and a stream stretch (42 58 0 33 00 N, 1 50 0 59 00 W) named "Alcoholera de Urritza," located upstream of the channelized area and without significant morphological alterations, which was used as a control (Fig. 1).This undisturbed site is close to the new channelized section (<1.5 km) and has no other significant impacts affecting it.Both stretches had similar habitat and morphological conditions before channelization.

Sampling Protocol and Data Acquisition
Sites were sampled eight times in the new channelized section and six times in the control from 1996 to 2020 (Table S1).The samplings were grouped according to successional stages: three samples for both sections in the short-term period (1996)(1997)(1998), two in the mid-term period (2005)(2006)(2007)(2008)(2009)(2010), three for channelized, and one for the control section in the long-term period (2018)(2019)(2020).At each sampling, water temperature ( C), dissolved oxygen (DO, mg/L), conductivity (μS/cm), and pH were measured with different monitoring systems (HI 98130,HI 9146 oximeter,and HI 98129 Combo Waterproof,Hanna Instruments, Woonsocket, RI, U.S.A.; WTW Multi 340i, Xylem Analytics Germany Sales GmbH & Co, Oberbayern, Germany).Habitat structures were characterized using transects established using the guidelines of Simonson et al. (1994).The data collected included mean width (m), depth (cm), water velocity (m/s), tree canopy shading percentage and dominant substrate categorized as fines (<2 mm), gravels ($2-64 mm), pebbles ($64-256 mm), boulders (>256 mm), and bedrock.These values were measured along perpendicular transects, regularly spaced crosssections of the channel.The number of and distance between transects were determined by dividing the total sampled reach length (100 m in length approx.)by 10.Along each transect, we measured the width (considering only the wetted surface), water depth, and water velocity.We determined whether the point was shaded and the type of substrate at 1-m intervals in the transect (Simonson et al. 1994).The diversity of depths was calculated by applying the Shannon diversity index to the different percentages of depth categories: low (<0.3 cm), moderate (30-60 cm), and high (>60 cm).Water velocity was measured with a rheometer (manufactured, boat display SpeedCoach OC 2, Nielsen-Kellerman Co., Boothwyn, PA, U.S.A.).
Fish sampling was conducted along a transect by electrofishing surveys with different back-pack electrofishers (Martín Pescador III, 300-600 V, 0.2-2 A, Acuitec SL, Aduna, Spain; Hans Grassl Model IG200 and IG200/2D, 300-600 V, 0.2-2 A, Hans Grassl GmbH, Schönau am Königsse, Germany), following a single-run methodology and estimating relative fish abundance based on catch per unit effort (CPUE; Meador et al. 2003).Principal habitat types were sampled in continuous wadable sampling transects (100 m in length).The collected fish were anesthetized and subsequently counted, their total length (TL to the nearest mm) was measured, and they were then released after the survey.Surveys were performed in accordance with the procedures set out in CEN standard EN 14011 (the European Standard for the sampling of fish with electricity) and with the corresponding authorizations from the local administrations in the different sampling campaigns.

Size Metrics
We determined the fish size spectra per stream sampling site and the time period following Arranz et al. (2021), who used a binned technique based on previous literature on aquatic food webs (Sprules & Barth 2016) to categorize fish by size into discrete bins or size classes and then analyzed the abundance of fish within each size class.The range of TLs of fish was between 16 and 515 mm.To classify the lengths of fish into 20 size classes to have enough data for the data analysis, we divided the range into a geometric series of 1.2 (length data to define body size), where size intervals become progressively wider with increasing body size.To normalize the fish abundance in each size class, we divided it by each size class width.To account for low electrofishing efficiency for the smallest stream fish (Bodine et al. 2011), we calculated the mode of the size classes containing the maximum abundance in all samples, which was the ninth size class (69-83 mm), and we started the size spectra from there.Before normalization, all individuals smaller than 69 mm (only representing 16% of the total fish size data) were accumulated in this ninth class.The linear size spectrum was calculated for each site and time period through ordinary leastsquare regression (OLS regression) between the log2 midpoint of size classes (independent variable) and the log2 of the total number per size class (dependent variable; Table S1; Fig. S1).Size diversity was calculated using the index described by Quintana et al. (2008), which considers the amplitude of the size range and the evenness across sizes.

Statistical Analysis
We used analysis of variance (ANOVA) and Bonferroni post hoc tests to compare the variance of the different habitats and fish metrics individually (dependent variables) across the different successional stages (short term, 1996-1998 years; mid term, 2005-2010 years; long term, 2018-2020 years) of the new channelized stream section (i.e.Corta de Urritza) and the control stream section (i.e.Alcoholera de Urritza).Habitat metrics included water temperature, DO, pH, depth, diversity of depths, water velocity, substrate, and shade.Fish metrics included the number of individuals, species richness, species Shannon diversity, mean length, maximum length, size diversity, slope of the size spectra, and intercept of the size spectra across the different successional stages (short term, 1996-1998; mid term, 2005-2010; and long term, 2018-2020) between the new channelized and control stream sections.When the original data did not satisfy the assumptions of normality (Shapiro-Wilks test) and homogeneity of variances (Levene's test), values were power transformed following the Box-Cox method (Box & Cox 1964).With this large number of tests performed, the Bonferroni correction controlled for false positives, at the cost of potentially increasing the risk of false negative, Type II errors (Nakagawa 2004).
In addition, we performed linear multiple regression models with the same variables but without grouping by successional stages to estimate: (1) the effect of the impact (independent variable 1) on the fish or habitat metric (dependent variable) to assess alterations between the impact and control site; (2) the effect of date (independent variable 2) on the fish or habitat metric (dependent variable) to evaluate changes over time; and (3) the significance of the interaction (impact* date) to determine whether there is a significant interaction effect between the fish or habitat metric, the impact, and time.
We used three principal component analyses (PCAs) to summarize variation among (1) substrate type (Fig. S2), (2) fish metrics, and (3) habitat metrics across the different successional stages (short, mid, and long term) between the new channelized stream section and the control stream sections.We used the first axis of these PCAs (those explaining the maximum variation and primarily related to the succession of the stream section after colonization) to evaluate how variation among the different successional stages and control stream section was explained by composite variation in substrate, fish, and habitat metrics.Data were analyzed using R v. 4.0.5 (R Core Team 2021).

Results
Most of the habitat variables changed during the succession process in the new channelized stream section until reaching values comparable to the control site in the long-term successional stage (Figs.2, S4).At the beginning (i.e.short-and mid-term successional stages), the mesohabitat of the new stream section was very simple and dominated by runs.Over the years that followed (i.e.long-term successional stage), the complexity increased, and mesohabitats became more diverse, with an increase in pools, depth, and diversity of depths.The substrate also evolved from pebbles and bedrock to boulders (Fig. S1), and the water temperature decreased in line with the increase in shade produced by the expansion of the riparian forest.All this variation along the successional stages in the new stream section was summarized by the first axis of the PCA for habitat variables from short-term to long-term succession stages, and the latter long-term succession samplings were located very near the control site values (explaining 54% of the variation, Fig. 3).The second axis (18% variability) pooled conductivity, water Restoration Ecology November 2023 temperature, and velocity.This second axis could be more related to natural environmental variability than to successional stages (Fig. 3).
In total, 2,263 fish were caught by electrofishing corresponding to 6 species (Barbatula quignardi, Gobio lozanoi, Luciobarbus graellsii, Parachondrostoma miegii, Phoxinus bigerri, and Salmo trutta).In the new channelized stream section, species richness and species diversity decreased over the years, from the presence of species found in the middle and lower reaches of rivers occurring in the area (such as Barbatula quignardi, Luciobarbus graellsii, or Phoxinus bigerri) to the unique dominance of Salmo trutta, as it was at the control site (Figs.4, S3).The mean and maximum lengths of the fish community in the new stream section increased through the succession process until reaching values comparable to those of the control site.Similarly, the slopes of the fish size spectra in the new channelized stream section also needed time to reach values comparable to the control site, indicating a transition from small fish to a broader representation of fish from all size classes in the long term (Table S1).Similarly, the size diversity of the fish community also increased over the years until comparable values were achieved with the control site.Although the abundance of fish individuals and carrying capacity of the system (i.e. the intercept of size spectra) were lower than those at the control site at the beginning of the colonization and succession (i.e.short-term and mid-term successional stages), later in the long-term period successional stage, they increased until acquiring equal, or even higher, values than in the control stream section (Table S1; Figs. 4, S1).These changes in fish metrics along the successional stages (from short-term to long-term successional stages) were very well reflected on the first axis of the PCA (explaining 69% of the variation; Fig. 5).We included linear multiple regression models as an additional analysis in the supplementary material (Tables S2,  S3; Figs.S3, S4).These results were consistent with the ANOVA findings.

Discussion
Natural resilience usually permits ecosystems (such as rivers) to change to withstand human disturbances and maintain ecological functions.When modified by artificial structures, flowing  (short-t, 1996-1998), mid-term (mid-t, 2005-2010), and long-term (long-t, 2018-2020) succession after 1995 when channelization was performed.Control refers to the metrics at an upstream river sampling site on the Larraun River in Alcoholera de Urritza without any impacts.Lowercase letters inside the plots refer to significant differences among successional stages (pairwise t-test with Bonferroni correction, p < 0.05).Each letter represents a distinct group or category with statistically significant differences.The big black circles represent the mean, the vertical lines represent AE SD and the small gray circles are the data points.Water temperature is in C, DO in mg/L, depth in cm, water velocity in m/s, and the substrate is represented by the scores of PCA1 (Fig. S2) and shade in percentage.*Water temperature values were power transformed to achieve normality of variances.
waters can recover their natural ecological status, showing a great recovery ability (White & Stromberg 2011).However, resilience has limits.Significant disturbances can undermine the ability to correct impacts, sometimes stopping them and sometimes slowing down the process.
We developed a long-term monitoring study on a new channelized stream, where no restoration actions were performed.Despite our dataset limitations, due to the relatively small sample size, the results clearly showed that 25 years later, the new channelized stream section had not yet reached the reference conditions of the control site.Although habitat complexity is currently high, some environmental characteristics are still changing.The riparian forest cover at the control site is almost complete, whereas the percent canopy cover is close to 50 in the new channelized stream section.The diversity of depths or substrate complexity can still achieve higher values in the new channelized stream section compared to the control site.
On the other hand, variables related to fish populations were close to those at the control site only in the long-term period (2018-2020), whereas principal mesohabitat changes were present in the mid-term period successional stage (2005)(2006)(2007)(2008)(2009)(2010).
Our results suggest that a long period is needed to achieve the reference conditions for fish assemblages.This fact is because mesohabitat features and habitat complexity strongly determine fish populations (Magalhães et al. 2002;Palmquist et al. 2021); therefore, it was first necessary for the habitat of the channelized stream section to be naturalized to recover the typical fish assemblages in this area (dominated by trout).For example, the influence of riparian forest cover on aquatic biota can be related to the control of thermal river conditions.Water temperature is considered one of the most relevant factors constraining river fish assemblages because it affects the growth, fecundity and metabolism of fish (Fuller et al. 2022).Brown trout, similar to other migratory salmonids, are susceptible to changes in water temperature.Our results show how the observed increasing domain of this species in the new channelized stream section is related to the decrease in temperature, which depends on the increase in the forest riparian cover that provides shade and reduces solar radiation over the water surface (Fuller et al. 2022).
Previous terrestrial studies have shown that species richness usually increases initially during the succession process but Figure 3. PCA to summarize the habitat metrics in Corta de Urritza along the new channelized stream section from short-term (short-t, 1996-1998), mid-term (mid-t, 2005-2010), and long-term (long-t, 2018-2020) succession after 1995 when channelization was performed.Control refers to the metrics at an upstream river sampling site on the Larraun River in Alcoholera de Urritza without significant impacts.In the boxplot, the values of PCA1 and PCA2 are represented.To ensure comparability among the different components in the biplot, the values in the PCA plot were scaled.In the boxplot, the values of PCA1 and PCA2 are represented.Lowercase letters inside the plots refer to significant differences among successional stages (pairwise t-test with Bonferroni correction, p < 0.05).Each letter represents a distinct group or category with statistically significant differences.Different letters indicate significant differences.Interquartile ranges (25th and 75th percentile) are shown by the height of the boxes, and horizontal lines represent medians (50th percentile).Whiskers range from the 10th to 90th percentiles.
often declines (Anderson 2007).Our results match this pattern because fish diversity and richness metrics decreased during the analyzed period to approximate the reference population in the new channelized stream section.Control and long-term channelized site samplings show very low richness and diversity, where brown trout is the dominant species, according to the usual fish assemblage in the upper stretches of rivers in the area (Leunda et al. 2012).We suggest that this trend of richness was a response to habitat changes.Throughout the years, the new channelized stream section has been naturalized.The habitat features (i.e.pools, depths) have become more suitable for trout and less suitable for the other species (Barbatula quignardi, Luciobarbus graellsii, or Phoxinus bigerri), which they likely have moved to the downstream reaches where environmental conditions are more appropriate for these species (Leunda et al. 2012).
The size-related variables of fish communities did show an increase during the fish succession process in the new channelized stream section.Size diversity showed a faster response than size spectra because it had comparable diversity with the control site in the mid-term successional stage.There was a transition from a low diversity of fish sizes (i.e.mostly small fish) just after the creation of the new stream section to a more size-structured fish community at mid-and long-term successional stages.On the other hand, fish size spectra in the new channelized stream section needed a more extended time period (i.e.long-term successional stage) to reach values comparable to the control site, that is, flatter slopes as the succession process progressed.Previous studies have shown that flatter slopes appear when fish communities have a high proportion of large fish (Emmrich et al. 2011;Arranz et al. 2021).This is consistent with the findings of our study, which showed that fish mean and maximum lengths increased over the years.Larger specimens need pools, a diversity of depths and boulders (Benejam et al. 2018), and these conditions only appeared in the new channelized stream section in the long-term successional stage.On the other hand, the carrying capacity of the community, represented by the abundance of fish individuals and the intercept of the size spectra, was low in the short term in the new channelized stream section.Nevertheless, it increased over the years and finally achieved values close to those of the control site.This late response of fish community carrying capacity is a consequence of fish dependence on lower trophic levels and suitable habitats (Magalhães et al. 2002).Therefore, basal river food webs must first be established and structured in the new habitat to allow the maintenance of future fish populations.Moreover, the carrying capacity of the river food web also depends on inputs of subsidies of allochthonous organic  -t, 1996-1998), mid-term (mid-t, 2005-2010), and long-term (long-t, 2018-2020) succession after 1995 when channelization was performed.Control refers to the metrics at an upstream river sampling site on the Larraun River in Alcoholera de Urritza without significant impacts.Lowercase letters inside the plots refer to significant differences among successional stages (pairwise t-test with Bonferroni correction, p < 0.05).Each letter represents a distinct group or category with statistically significant differences.Different letters indicate significant differences.The big black circles represent the mean, the vertical lines represent AE SD and the small gray circles are the data points.Lengths are in mm.*The number of individuals, slope, and intercept of the size spectra were power transformed to achieve normality of variances.
Fish community recovery on channelized stream carbon (Benejam et al. 2018), and this could appear when riparian forests have been developed in the new channelized stream section (i.e.long-term successional stage).
Overall, our results indicated that a long-term period was needed to naturalize the new channelized stream section.However, some restoration procedures might have significantly reduced this period (Palmquist et al. 2021).Fortunately, during the last two decades, in the context of new environmental laws and requirements, many river sections have been restored to return these ecosystems to their most natural state possible or an equivalent state (White & Stromberg 2011).These projects provide suitable lessons on how to recreate habitats and improve biodiversity and ecosystem services to reverse the environmental degradation caused by human activities (Yang et al. 2017).Improvements in the riparian forest (i.e.planting of native shrubs and trees), bank shaping and reinforcement (i.e. the inclusion of live willow fascine and stakes), and riverbed restoration (i.e.including diverse sizes of substrates from gravel to large boulders and large woody debris) have been shown to be efficient measures to restore fluvial ecosystems (Zika & Peter 2002;Pretty et al. 2003).This kind of small-scale restoration scheme is effective when nearby natural reaches are sustaining enough potential colonists (Pretty et al. 2003).Steep and deeper banks provide cover for larger fish, especially if abundant overhanging vegetation is also available.In the same way, the introduction of large woody debris can serve to improve salmonid habitat (Zika & Peter 2002).Several studies have suggested that accounting for habitat heterogeneity greatly increases restoration efficiency and should be considered when planning these activities (Yang et al. 2017).
In conclusion, constraining environmental features significantly shape the fish community and distribution of freshwater fish species in the colonization and succession process.The results of the present study suggest that habitat complexity (diversity of substrates, depths, etc.) is correlated to the structure of fish populations, and small-scale restoration measures should determine the colonization process, enabling the natural reference conditions to be reached earlier.The size spectra and other size-related variables changed systematically in the new channelized stream section over time.They therefore may be good ecological indicators when assessing the evolution of fish colonization and succession.Acquiring a 25-year time series of fish data, as in our study, has an important complexity associated.However, considering that we have detected that changes in habitat and fish communities occur at a very slow rate, our relatively small sample size (due to a low sampling frequency within the Figure 5. PCA to summarize the fish metrics in Corta de Urritza along the new channelized stream section from short-term (short-t, 1996-1998), mid-term (mid-t, 2005-2010), and long-term (long-t, 2018-2020) succession after 1995 when channelization was performed.Control refers to the metrics at an upstream river sampling site on the Larraun River in Alcoholera de Urritza without significant impacts.In the boxplot, PCA1 and PCA2 are represented.Lowercase letters inside the plots refer to significant differences among successional stages (pairwise t-test with Bonferroni correction, p < 0.05).Each letter represents a distinct group or category with statistically significant differences.Different letters indicate significant differences.Interquartile ranges (25th and 75th percentile) are shown by the height of the boxes, and horizontal lines represent medians (50th percentile).Whiskers range from the 10th to 90th percentiles.
time series) may affect the generalizability of our findings.Future work should aim at obtaining longer time series and more frequent sampling.

Figure 1 .
Figure 1.Map with the location of the sampling sites (1: Alcoholera de Urritza, control site; 2: Corta de Urritza, the new channelized stream section) along the middle Larraun River.In the upper part of the figure, the location of Navarre on the Iberian Peninsula (left) and the Larraun River in Navarre (right).At the bottom of the figure, pictures of the new channelized stream section during the sampling period are shown.

Figure 2 .
Figure2.Variability of habitat metrics in Corta de Urritza, the new channelized stream section from short-term(short-t, 1996-1998), mid-term (mid-t, 2005-2010), and long-term (long-t, 2018-2020) succession after 1995 when channelization was performed.Control refers to the metrics at an upstream river sampling site on the Larraun River in Alcoholera de Urritza without any impacts.Lowercase letters inside the plots refer to significant differences among successional stages (pairwise t-test with Bonferroni correction, p < 0.05).Each letter represents a distinct group or category with statistically significant differences.The big black circles represent the mean, the vertical lines represent AE SD and the small gray circles are the data points.Water temperature is in C, DO in mg/L, depth in cm, water velocity in m/s, and the substrate is represented by the scores of PCA1 (Fig.S2) and shade in percentage.*Water temperature values were power transformed to achieve normality of variances.

Figure 4 .
Figure 4. Variability of fish metrics in Corta de Urritza along the new channelized stream section from short-term(short-t, 1996-1998), mid-term (mid-t, 2005- 2010), and long-term (long-t, 2018-2020) succession after 1995 when channelization was performed.Control refers to the metrics at an upstream river sampling site on the Larraun River in Alcoholera de Urritza without significant impacts.Lowercase letters inside the plots refer to significant differences among successional stages (pairwise t-test with Bonferroni correction, p < 0.05).Each letter represents a distinct group or category with statistically significant differences.Different letters indicate significant differences.The big black circles represent the mean, the vertical lines represent AE SD and the small gray circles are the data points.Lengths are in mm.*The number of individuals, slope, and intercept of the size spectra were power transformed to achieve normality of variances.